Working Paper: NBER ID: w2556
Authors: Kurt Beron; Helen V. Tauchen; Ann Dryden Witte
Abstract: In this paper. we estimate a three equation model for taxpayers' reported income and tax liability and for the probability of an audit. Our work differs from previous studies in that our dependent variables in the compliance equations are taxpayer reports rather than a variable related to auditor estimates of noncompliance and in that we estimate a structural equation for audits. We find that audits stimulate compliance although the effect is not large and is not statistically significant for all groups. Audits are more effective at inducing accurate reporting of subtractions from income than of income. Reduced-form results suggest that IRS activities other than audits have significant compliance effects. Results for the sociodemographic variables are interesting and help to explain some seemingly incongruous findings in the literature. We find compliance to be higher, if anything. in areas with less educated and older taxpayers, a large proportion of households headed by females. and a mostly native born population.
Keywords: Tax Compliance; Auditing; Structural Equation Model
JEL Codes: H26; K34
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| audits (M42) | compliance (K40) |
| audits (M42) | reported adjusted gross income (AGI) (E10) |
| audits (M42) | tax liability (H20) |
| IRS activities beyond audits (H26) | compliance (K40) |
| audit rates (M42) | reported AGI (E10) |
| audit rates (M42) | tax liability (H20) |